pages/review_predictor.py

import base64
import json
import pickle

import pandas as pd
import streamlit as st
import torch
import torch.nn as nn
import transformers

from model.funcs import (create_model_and_tokenizer, execution_time,
                         load_model, predict_sentiment)
from model.model import LSTMConcatAttentionEmbed
from preprocessing.preprocessing import data_preprocessing
from preprocessing.rnn_preprocessing import preprocess_single_string


def get_base64(file_path):
    with open(file_path, "rb") as file:
        base64_bytes = base64.b64encode(file.read())
        base64_string = base64_bytes.decode("utf-8")
    return base64_string


def set_background(png_file):
    bin_str = get_base64(png_file)
    page_bg_img = (
        """
    <style>
    .stApp {
    background-image: url("data:image/png;base64,%s");
    background-size: auto;
    }
    </style>
    """
        % bin_str
    )
    st.markdown(page_bg_img, unsafe_allow_html=True)


set_background("main_background.png")


@st.cache_resource
def load_logreg():
    with open("vectorizer.pkl", "rb") as f:
        logreg_vectorizer = pickle.load(f)

    with open("logreg_model.pkl", "rb") as f:
        logreg_predictor = pickle.load(f)
    return logreg_vectorizer, logreg_predictor


logreg_vectorizer, logreg_predictor = load_logreg()


@st.cache_resource
def load_lstm():
    with open("model/vocab.json", "r") as f:
        vocab_to_int = json.load(f)

    with open("model/int_vocab.json", "r") as f:
        int_to_vocab = json.load(f)
    model_concat_embed = LSTMConcatAttentionEmbed()
    model_concat_embed.load_state_dict(torch.load("model/model_weights.pt"))

    return vocab_to_int, int_to_vocab, model_concat_embed


vocab_to_int, int_to_vocab, model_concat_embed = load_lstm()


@st.cache_resource
def load_bert():
    model_class = transformers.AutoModel
    tokenizer_class = transformers.AutoTokenizer
    pretrained_weights = "cointegrated/rubert-tiny2"
    weights_path = "model/best_bert_weights.pth"
    model = load_model(model_class, pretrained_weights, weights_path)
    tokenizer = tokenizer_class.from_pretrained(pretrained_weights)

    return model, tokenizer


model, tokenizer = load_bert()


@execution_time
def plot_and_predict(review: str, SEQ_LEN: int, model: nn.Module):
    inp = preprocess_single_string(review, SEQ_LEN, vocab_to_int)
    model.eval()
    with torch.inference_mode():
        pred, _ = model(inp.long().unsqueeze(0))
    pred = pred.sigmoid().item()
    return 1 if pred > 0.75 else 0


def preprocess_text_logreg(text):
    # Apply preprocessing steps (cleaning, tokenization, vectorization)
    clean_text = data_preprocessing(
        text
    )  # Assuming data_preprocessing is your preprocessing function
    vectorized_text = logreg_vectorizer.transform([" ".join(clean_text)])
    return vectorized_text


# Define function for making predictions
@execution_time
def predict_sentiment_logreg(text):
    # Preprocess input text
    processed_text = preprocess_text_logreg(text)
    # Make prediction
    prediction = logreg_predictor.predict(processed_text)
    return prediction


metrics = {
    "Models": ["Logistic Regression", "LSTM + attention", "ruBERTtiny2"],
    "f1-macro score": [0.94376, 0.93317, 0.94070],
}


df = pd.DataFrame(metrics)
df.set_index("Models", inplace=True)
df.index.name = "Model"


st.sidebar.title("Model Selection")
model_type = st.sidebar.radio("Select Model Type", ["Classic ML", "LSTM", "BERT"])


styled_text = """
<style>
.styled-title {
    color: #FF00FF; 
    font-size: 40px;
    text-shadow: -2px -2px 4px #000000;
    -webkit-text-stroke-width: 1px;
    -webkit-text-stroke-color: #000000;
}
.positive {
    color: #00FF00; 
    font-size: 30px;
    text-shadow: -2px -2px 4px #000000;
    -webkit-text-stroke-width: 1px;
    -webkit-text-stroke-color: #000000;

}
.negative {
    color: #FF0000; 
    font-size: 30px;
    text-shadow: -2px -2px 4px #000000;
    -webkit-text-stroke-width: 1px;
    -webkit-text-stroke-color: #000000;

}
</style>
"""

st.markdown(styled_text, unsafe_allow_html=True)

# Streamlit app code
st.markdown('<div class="styled-title">Review Prediction</div>', unsafe_allow_html=True)
text_input = st.text_input("Enter your review:")
if st.button("Predict"):
    if model_type == "Classic ML":
        prediction = predict_sentiment_logreg(text_input)
    elif model_type == "LSTM":
        prediction = plot_and_predict(
            review=text_input, SEQ_LEN=25, model=model_concat_embed
        )
    elif model_type == "BERT":
        prediction = predict_sentiment(text_input, model, tokenizer, "cpu")

    # Apply different styles based on prediction result
    if prediction == 1:
        st.markdown(
            f'<div class="positive">Отзыв положительный</div>', unsafe_allow_html=True
        )
    elif prediction == 0:
        st.markdown(
            f'<div class="negative">Отзыв отрицательный</div>', unsafe_allow_html=True
        )

st.write(df)